Machine Learning for Structural Integrity Assessments

The continued safe operation of critical infrastructure is key in ensuring economic prosperity.

The drive towards carbon neutrality presents significant challenges to engineers as increasing operational efficiency often results in harsh, unfavourable operating conditions (e.g. offshore wind turbines).

The cost of implementing more efficient processes is a reduction in materials performance resulting in prohibitively short component lifetimes. The challenge facing engineers lies in improving the structural integrity assessment (SIA) methods.

Current SIA methods are predominately stress-based and thus, inherently dominated by the yield strength of the component material.

Non-linear materials such as steels typically fail by strain induced plasticity where significant additional energies are adsorbed prior to fracture. Strain-based assessments contain considerable built in conservatisms that have not yet been explored.

The principal aims of this fellowship application are to develop more advanced SIA methods by considering conservatisms in existing stress-based and strain-based approaches and to exploit recent advances in machine learning to identify and predict key parameters influencing transformative damage in fracture toughness testing.

The fundamental understanding of material damage generated in this work will reduce knowledge gaps currently impeding SIA improvement.

The benefits of this work include advances to multiple international codes and standards, the continued safe operation of aging critical infrastructure, longer more realistic estimated lifetimes for new components and, significant industrial cost savings through enhanced component design, reduced maintenance cost and reduction in early structure retirement.